Hi all,
I read with interest Bruce's post that he wants to work on another
Linux distribution. :-)
As long as we are talking "pie in the sky" stuff, I thought I'd let
loose with the news that I am also developing an alternative Linux
distribution. I've sort of hinted about it on several of my webpages
anyways. I've had this idea for two years (at least), but it's just
lately that I have actually started to work on it.
Here are my plans:
- mid/late-1999 for first release (codename: Skaha) - I'm not rushing
things. :-)
- it will be a strict subset of Debian (and will include support for
upgrading to full Debian distribution via apt) - basically, Debian
is the "upstream" distribution, and this would be a "downstream
distribution"
- dpkg based
- probably using apt (perhaps with an additional front-end)
- targetted towards desktop use only, no server apps, just a few games
- minimal size - optimized for installation via 28.8k modem via FTP,
which will be the primary distribution mechanism (not CD).
- slickest, easiest install in the business
- self-hosting (it has everything needed to compile itself)
- 100% open source / DFSG compliant (no non-free or contrib packages)
- minimal set of install options:
1) console only (with base application set)
2) X11 user (primarily Gnome/Gtk based, I hope)
Includes base application set, with:
one window manager, one editor, one file manager,
one email/news program, several web browsers (with one
front end), one documentation system, several scripting
languages
The following application sets can be chosen (preliminary):
2a) graphics - Gimp, 3D apps
2b) publishing - TeX, SGML, etc.
2c) small business - GnuCash, Inventory, Taxes
2d) CAD/Robotics/Control Systems
3) Developer - all tools and libraries required to make distribution
self hosting.
- sysadmin (and users) can use apt to install additional packages from
Debian distribution.
- initial distribution will be i386 GNU/Linux, localized for Canada
and the USA. Beyond that, I may attempt a Cygnus GNU/Win32 hosted
distribution, and maybe ever a GNU Hurd based one.
- very strong technical direction. As little duplication as possible in
terms of applications and libraries. Everything will be document-centric,
and tutorial driven. There will be a single document model.
It will be another Linux distribution - but it will also be a strict
subset of stuff inside Debian. It will be promoted as such.
It will not have an "open" development model as Debian has. Actually,
initially, it will only have one developer (me). If my consulting
business improves, I might consider paying people to help me out with
it. I don't want any volunteers however, as that would lessen my
control (anybody volunteering for Debian will be helping me out
anyways).
Most of the work I do will be "open" though, because it will appear in
Debian first.
The best way to think of this project of mine is basically as "Debian
+ editorial control". Or you might think of it as a "Jim's hobby
project" -- a customized "hot rod" GNU/Linux.
If it turns out well, I'll consider marketing it commercially (it
should be marketable). I will offer a consulting service that
consists of customizing the distribution for clients with
"stylesheets" and "themes". I imagine that many local ISPs might want
to offer their own customized version of Linux to their local small
business clients and home users.
As it will be "Open Source", other more-commercial distributions such
as Red Hat, Caldera and SuSe can steal stuff with no complaints from
me.
The driving motivation for doing this isn't commercial though - it's
basically a "soapbox" for me to express what I think should be in an
OS, and what shouldn't. My point of view is that most of the code
that a user really needs has been written already - most of the work
needed to build a really nice OS doesn't consist of writing more code
(too many free software projects are going down this road). Mostly,
it's just a packaging/integration and documentation job to make the
current codebase more digestable. Debian already does a lot of this,
but this will be focused in a way Debian can't be.
I plan to develop some documentation and live tutorial technology
which will go far beyond what any free software app currently has.
The whole system will be keyed off of the documentation - rather than
leaving it as an afterthought. Much of the work I'm going to put into
preparing documents and metadata for LinuxHQ and dwww will also go
into this new system.
The release schedule will vary dramatically from Debian's. There will
be stable and development branches. There will be a new stable
release approximately every two months (built from packages from the
Debian unstable distribution, but tested). Security bugs will be
immediately put into the stable release.
Because there will only be a small set of packages to test, and a
small closely-knit release team, rapid stable releases can be done.
For the most part, the development branch will just be a strict subset
of the Debian unstable distribution. Basically, it will be the same
packages. Due to the differing release schedule, there will probably
need to be quite a few non-maintainer fixes - but these will all be
pushed upstream to Debian. Any packages tha I develop specifically
for this distribution, I will also upload to Debian.
There will be a separate bug system, but most bugs will be fed
"upstream" to the Debian bug system (with patches if they are easy
fixes). There will be relatively few bugs, because the packages
released into stable will have been rigorously tested. There will be
separate mailing lists in addition to the Debian mailing lists.
Unlike Bruce's project - this project will have a symbiotic
relationship with Debian. If it works, it will attract new users to
the Debian code base. Also, the more successful Debian is, the more
successful this subset of Debian will be.
I'd like to see more people announce that they want to develop their
own "subset" Linux distributions based on Debian. I'd be willing to
collaborate on tools to make this easier.
The big problem with Bruce's idea of developing yet another volunteer
distribution is that he will once again have no control over what the
volunteers will be doing. He'd be much better off using the same
model I am going to use, which allows for near 100% editorial control,
without giving up the benefits of having hundreds of developers
feeding code in. Of course, if he wants to base a distribution on rpm
rather than dpkg (bad idea, IMHO), he'd be better off basing his work
on the Stampede distribution, and organizing a recruiting campaign for
them.
In summary, don't hold your breath for my "subset distribution" of
Debian - it will take a long time to pull together. But I strongly
believe that this is a model Debian should encourage.
The Debian distribution "proper" may never have more market share than
the commercial distributions such as Red Hat, Caldera, and SuSe.
However, it is entirely possible that derivative subset distributions
of Debian could dominate the Linux marketplace (especially given the
technical superiority of Debian).
Cheers,
- Jim
Attachment:
pgpMVd9liqoa0.pgp
Description: PGP signature